Scholarlink

<p>Resilient service composition in cloud networks is essential for ensuring continuous and reliable service delivery amid failures and dynamic conditions. Traditional approaches often face scalability and adaptability challenges, particularly when implementing complex resilience strategies like load balancing and anti-affinity rules. In this study, a novel Quantum Reinforcement Learning (QRL) framework is proposed to enhance cloud service resilience. Leveraging quantum computational principles via the PennyLane framework, the QRL agent effectively learns optimal task allocation strategies, integrating anti-affinity constraints to mitigate single points of failure and enhance robustness. The approach is validated against classical reinforcement learning (RL) methods through a comprehensive case study. Results demonstrate that the QRL agent achieves significant improvements, reducing training time by approximately 29% and memory usage by 51% compared to the classical RL baseline, while also generating more diverse and resilient service compositions. These quantitative enhancements underline the potential of quantum computational intelligence for efficient, scalable, and resilient cloud service composition.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Quantum reinforcement learning for resilient cloud service composition

  • 期刊论文

  • 发表日期：

    2026年1月27日

查看全文
  • Erfan Shahab,
  • Sharareh Taghipour

摘要

Resilient service composition in cloud networks is essential for ensuring continuous and reliable service delivery amid failures and dynamic conditions. Traditional approaches often face scalability and adaptability challenges, particularly when implementing complex resilience strategies like load balancing and anti-affinity rules. In this study, a novel Quantum Reinforcement Learning (QRL) framework is proposed to enhance cloud service resilience. Leveraging quantum computational principles via the PennyLane framework, the QRL agent effectively learns optimal task allocation strategies, integrating anti-affinity constraints to mitigate single points of failure and enhance robustness. The approach is validated against classical reinforcement learning (RL) methods through a comprehensive case study. Results demonstrate that the QRL agent achieves significant improvements, reducing training time by approximately 29% and memory usage by 51% compared to the classical RL baseline, while also generating more diverse and resilient service compositions. These quantitative enhancements underline the potential of quantum computational intelligence for efficient, scalable, and resilient cloud service composition.